ERS Scatterometer Land Applications: Detecting Soil Thawing in Siberia

K. Boehnke & V. R. Wismann

Institute for Applied Remote Sensing,
Bahnhofstrasse 54, 22880 Wedel, Germany
Phone: +49-4103-13922 Fax: +49-4103-7469
E- mail: 100412.1731@compuserve.com

Based on multitemporal, multiple incidence angle radar backscatter measurements over Siberia, obtained from the ERS-1 scatterometer, a method has been developed for monitoring the extent of frozen soils. Throughout the course of the year, the normalised radar cross-section (NRCS) data show distinct variations which can be attributed to the thawing and freezing of soils.

In spring, during snow melt, the NRCS first decreases by up to 5 dB and shortly later, when the soils start thawing, increases dramatically. Due to the huge extent of the Siberian test site, it takes more than three months for the entire region to be thawed. For the years 1992 to 1995, maps of the isochrons of thawing were constructed for the Siberian test site which reveal the geographical distribution and the interannual variability of the 'onset of thawing'. Overall, thawing commences in the south/southwest and progresses towards northern central Siberia. The thawing of mountainous regions is delayed compared to regions of lower elevation.

Introduction

The C-band (5.3 GHz) scatterometer of the ERS-1 satellite operates with vertical polarisation, both for transmission and reception (VV). Its incidence angle ranges from 18 to 57°, the illuminated swath is 500 km wide, and the along-track and cross-track spatial resolution is 50 km (ESA, 1992). The scatterometer measurements are independent of cloud coverage and illumination by the sun and, therefore, superior to measurements by optical systems. Compared to the Synthetic Aperture Radar (SAR) aboard ERS-1, the scatterometer delivers a manageable amount of data (global coverage within 3 to 4 days) and a geometric resolution which is reasonable for many applications (Wismann et al., 1996).

When the vegetation cover is sparse or absent, the normalised radar cross-section (NRCS) at C-band, VV polarisation, depends mainly on the moisture content of the soil, the dielectric constant, the penetration depth and the surface roughness. The radar backscatter increases with soil moisture and decreases with surface roughness. When liquid water in soil freezes, the dielectric constant of the soil falls dramatically. This process is reversed in uspring when the soil thaws. Figure 1 shows the dependence of the NRCS at 40°-57° incidence angle on the air temperature as measured at the World Meteorological Organisation (WMO) station Kolpashevo (58.2°N, 82.9°E) (Deutscher Wetterdienst, 1991-1994). A step-like increase in NRCS of approx. 2 dB can be seen when the monthly mean air temperature exceeds 0°C.

Figure 1. NRCS(40-57) versus air temperature (monthly means) at the WMO station Kolpashevo (58.2°N, 82.9°E).

Information on the state of the soils (frozen/thawed) is of great importance for climate modelling because evapo-transpiration and gas exchange processes change significantly when going from one state to the other. Also, practical applications (transportation, construction and oil/gas production) would benefit from near- real-time estimates of the extent of frozen ground.

Soil State

Detection In this investigation, ERS-1 scatterometer data distributed on CD-ROM by CERSAT, the French Processing and Archiving Facility, were used. These data are superior in quantity to the GTS or FDC distributions, where a large part of the Siberian test site is lacking data from descending orbits due to delayed transmission of the data to the distribution centres.

Detecting the thawing of the soils requires two steps: first, the NRCS data are re-sampled to a grid with a resolution of 0.5° in latitude and 1° in longitude, which corresponds roughly to the instrument resolution of 50 50 km. For three-day intervals, an average radar cross-section (NRCS-40) is computed for each grid point by linear regression between the measured NRCS and their respective incidence angles. NRCS-40 is the value of this regression line at an incidence angle of 40°. This procedure accounts for the incidence angle dependence of the NRCS, which is highly variable in spring when the backscattering mechanism changes from volume to surface scattering depending on snow cover, snow wetness, soil moisture and the penetration depth of the microwaves. Moreover, irregular distributions of NRCS measurements with respect to the incidence angle due to sampling and orbital characteristics have no influence on NRCS-40. Figure 2 shows two typical time series of NRCS-40 for two grid points in Siberia (68.75°N, 79.5°E and 56.25°N, 62.5°E, respectively). Note the extreme stability of the signal in early spring and the large variations (=5 dB) associated with the onset of thawing.

Figure 2. Time series of NRCS-40 for two grid points in 1993.

The second step involves determining the monthly means for February and July of each year and, finally, detecting the moment of thawing. The latter is achieved by analysing the NRCS-40 time series during the spring (for the time period 1 March to 1 July) for the step-like increase seen in Figure 1. A bin is marked as thawed when two consecutive NRCS-40 values exceed 50% of the difference between the monthly mean of NRCS-40 for July and February of the respective year. The constraint of requiring two values above the threshold eliminates spikes which are not removed when computing NRCS-40 and single events which are not connected to the thawing process.

In Figure 2, for the grid point located in the north, the onset of thawing is detected at the beginning of June, whereas further south, the algorithm detects thawing at the beginning of April. As can be seen from the NRCS-40 time series of the northern grid point, spring snow melting leads to a strong short-term decrease of up to 5 dB. This drop in NRCS-40 can be explained by the increasing wetness of the snow leading to an enhanced absorption of the microwaves and the subsequent low radar return from melt water ponds formed on the frozen grounds when infiltration is still blocked.

By applying the above algorithm to the Siberian test site (50- 80°N; 42-172°E), maps of the temporal change in the state of the soils (frozen to thawed) have been constructed for the years 1992-1995. Figures 3 and 4 show these maps for 1992 and 1995, respectively. The onset of thawing differs significantly in terms of time period and geographical evolution for these two years, especially in the south. In these maps, each colour represents the area thawed before the date marked under the colour scale. Each of the 16 colour codes correspond to a time step of 8 days and thus span the observation period from 1 March to 28 June. Superimposed are contour lines marking elevation levels of 100, 250, 500 and 1000 m (Lee & Hastings, 1995).

Figure 3. Map of the Siberian test site (50-80°N, 42- 172°E) depicting isochrons of the freeze/thaw transition of soils, deduced from ERS-1 scatterometer data for 1992. A region displayed in a certain colour has thawed before the period indicated under the colour scale. Superimposed are contour lines indicating elevation levels of 100, 250, 500 and 1000 m. The spatial resolution is 0.5° in latitude and 1° in longitude.

Figure 4. Same as Figure 3 but for 1995.

In 1992, the onset of thawing is late. After the southern part of the test site starts thawing in April, thawing gradually moves towards the north coincident over the complete range of longitudes (Fig. 3). By the middle of May, thawing has reached most of Siberia with the exception of the Central Siberian Plateau. For 1995 (Fig. 4), thawing commences early in the southwest and progresses towards the northeast somewhat slower than in 1992, so that by the middle of May in both years approximately the same area has thawed (green to grey colours). Finally, in both years, the Central Siberian Plateau starts thawing in June. On a regional scale, the onset of thawing follows orographic features. As expected, mountainous regions (e.g. Ural, Werchojansk, Stanowoi) thaw later than regions of lower elevation.

We are aware of the weakness of our definition of the point of time of thawing. The accuracy with which this point can be detected depends also on the temporal resolution of the in-situ data, and so far only monthly mean values of air temperature have been available. Future evaluation of more detailed in-situ data and SAR imagery (as shown by Rignot & Way, 1994) will clarify this point. But as one can see from Figure 2, any other definition of the point of time of thawing can only lead to a relative time shift of a few days.

Interannual Variability

For each time step of 3 days, the area classified as thawed was calculated for a sub-region (55- 70°N, 50-120°E) of the Siberian test site encompassing the West Siberian Lowland and the Central Siberian Plateau (Fig. 5). In 1993 and 1994, the area thawed gradually increases throughout the entire time period, whereas in 1992 the onset of thawing is late, but then progresses much faster than in the other years. In 1995 thawing begins very early and by the end of April covers a large area. Then thawing slows down, so that for all years the entire region is thawed at approximately the same time. These large-scale characterisations of 'Spring' for the years 1992 to 1995 agree with reported weather conditions (UN- WMO, 1992-95). From monthly snow cover data derived from SSM/I measurements (Grody, 1991), the late onset of Spring in 1992 and the early but slower progress in 1995 are consistent with the signatures found in the ERS-1 scatterometer data (Boehnke & Wismann, 1996).

Figure 5. The area detected as thawed as a function of time deduced from ERS-1 scatterometer data for the years 1992 to 1995.

Conclusion

A method has been developed for monitoring the state of soils from Arctic to temporate climate regions. It has been demonstrated that the proposed algorithm reveals reasonable geographical distributions and temporal evolution of the thawing of the soils. In general, thawing commences in the south and then progresses towards the north. Northern central Siberia is the last area to be thawed. Regions of higher elevation thaw later than lower regions. Interannual variations detected from evaluation of the scatterometer data are significant and agree well with those detected in SSM/I-derived snow cover data and reported weather conditions.

An advantage of the proposed method not mentioned so far is that, with minor modifications, the algorithm detecting the time of thawing is suitable for operational use and, thus, could make soil-state information available in near-real-time. This would provide a good data basis for climate models which are very sensitive to changes in the evapo-transpiration exchanges between the atmosphere and the ground, which essentially depend on the soil state.

Acknowledgments

This work was funded by the European Space Agency within the Earth Observation Preparatory Programme (EOPP) under ESTEC Study Contract 11103/94/NL/CN. We thank Chung-Chin Lin of ESTEC for his support during this study and Ralph Ferraro, Microwave Sensing Group, NOAA/ Satellite Research Laboratory, for provision of the SSM/I-derived snow-cover data.

References

Boehnke K & V Wismann, Thawing of soils in Siberia observed by the ERS-1 scatterometer between 1992 and 1995, submitted for publication in: Proc. Intl Geoscience & Remote Sensing Symp. IGARSS'96, Lincoln, USA, May 27-31, 1996

Deutscher Wetterdienst, Die Witterung in Ubersee, Seewetteramt Hamburg, ISSN 0043-7085, Vol.40, No.1; Vol.42, No.11, 1991-1994.

ESA, ERS-1 System, ESA Publ. Div. ESTEC Noordwijk, The Netherlands (ESA SP-1146), 87 p., 1992.

Grody NC, Classification of snow cover and precipitation using the SSM/I, J. Geophys. Res., 96, 7423-7435, 1991.

Lee WR III & D Hastings, TerrainBase Global DTM Version 1.0 (on CD-ROM), Nat. Geophys. Data Center & World Data Center-A for Solid Earth Geophys. Boulder, Colorado, USA, April 1995.

Rignot E & Way JB, Monitoring freeze-thaw cycles along North- South Alaskan Transects using ERS-1 SAR, Remote Sens. Environ., 49, 131-137, 1994.
United Nations, WMO, Climate System Monitoring Programme, Geneva, Monthly Bull., Issues No. 1-1992 to 9-1995.

Wismann V, A Cavanie, D Hoekman, I Woodhouse, K Boehnke & C Schmullius, Land surface observations using the ERS-1 scatterometer, Report for ESA, ESTEC Contract 11103/94/NL/CN 57 p., Feb. 1996.